Investigation Into Single Pixel Imaging: Images Without A Camera

• Main Author: Au, Zhishan
• Format: Final Year Project / Dissertation / Thesis
• Published: 2020
• Subjects: TJ Mechanical engineering and machinery
• Online Access: http://eprints.utar.edu.my/4224/1/1503203_FYP_report_%2D_ZHISHAN_AU.pdf; http://eprints.utar.edu.my/4224/

Over the past decades, CCD and CMOS have been dominating in the imaging sensor technology. Due to the cost and technological constraints especially in the unusual wavelengths and low light condition, single pixel imaging becomes an important alternative. This approach samples the target scene with a set of microstructured light masks and obtain the spatial information using only a simple photodiode as the detector. Single pixel imaging strongly depends on the correlation of the mask patterns with the target scene. In this research, various types of mask patterns were studied on their behaviours when sampled with different images. This project conducts a simulation study on single pixel imaging, analyses various mask patterns and reviews their results to form a conclusion. Masks can be separated into 2 categories, Deterministic (Fourier and Hadamard) and Pseudo-random (Random and Gaussian) masks. As summarized from the results, Random mask has the shortest masking time while Hadamard Mask shows the shortest image reconstruction time. In general, Pseudo-random masks are better for images with random details throughout the images, meanwhile Deterministic masks outperform for simple images. However, Deterministic masks can be customized to - give better performance at sampling certain random images. Furthermore, Deterministic masks can be pre-loaded and the imaging results are 100% reproducible while it’s not possible for Pseudo-random masks. In conclusion, Deterministic masks are overall preferred as compared to Pseudo-random masks. This in-depth investigation establishes a comprehensive understanding of the mask patterns and their influence on the performance of single pixel imaging.